20 Oct 01



Gun Death Rate Histories for States
Figure 1. Gun Injury Death Rate Histories for States/Territories
First, the death numbers for Tasmania were so low that they and the corresponding rates varied wildly from year to year except for '86, '87 & '88.  However, a moving average of two years at a time does result in a reasonably smooth plot that happens to be close to the same shape as the plot for Queensland, but with levels about 20-40 percent higher, as shown in Figure 1.

These two states were widely thought (apparently without proof) to be the most liberal as far as firearm availability was concerned.  The similarity of the two curves suggests that maybe the same things were responsible for the main variations in the gun death rates in both the states.  Note that the nearly continuous decline for Queensland started the year ('91) its Firearms Act of 1990 went into effect but the similar decline for Tasmania started in '92, a year before the effectivity ('93) of the Tasmania Guns Act 1991.  This suggests the decline for Tasmania, hence the one for Queensland, was caused by something other than the gun laws.

Tasmania Suicide Rates After Effectivity of Guns Act '91
Figure 2. Tasmania Suicide Rates Before & After
Effectivity of Guns Act '91
A check of gun death rates at the National Injury Surveilance Unit (NISU) at Flinders Univ. reveals that almost all of the gun death rates for Tasmania result from the gun suicide rates, which also establish almost exclusively the shape of the gun death rates plot.  A plot of suicide rates for '91 through '94 looks almost identical (in shape) to a plot for suicide by all methods other than firearm, and not at all like a plot of firearm suicide rates, which remained relatively constant.  The small reduction of the gun suicide rate from the '92 rate was a result only of the fact that the '92 rate was one of the random high rates found in the Tasmania data, indicating that the drop for '93 was just a return to more normal levels rather than anything to do with firearms laws.

The shape of the curve for Queensland is also nearly dictated by the gun suicide rates over the period covered by the graph except for a moderate increase of homicide in '86 along with small drops in homicide for '88, '93 & '94.  The small departures from the suicide curve shapes are not common between Tasmania and Queensland .  This means that the only commonality between the rates for the two states is in the area of gun suicides.  So, was there anything common between the two states in these years that would be likely to affect suicide rates?

Note (Figure 1) that Western Australia consistently had the lowest overall gun death rate that was relatively constant at about 2.5 to 3.3 from '83 to '94.  The rate for South Australia was relatively stable at about 3.5 from '83 through '89.  In '90 the rate started rising gradually to a high of about 4.4 in '92, but turned sharply down for '93 then rose back a little to 3.27 for 1994.


Figure 3. Gun Injury Death Rate Histories for NSW & Victoria
The curve for NSW is repeated in Figure 3 along with the plot for Victoria.  The gun death rate history for Victoria was unique in having the prominent peak for '87 and '88 even though the rate generally fell throughout the rest of the '83-'94 period.  The curves for three of the other states had abrupt changes in direction at '87; but the change was the opposite for W.A. as it was for Queensland and Tasmania, so the idea that something that caused the major peak in Victoria also caused lesser changes in other states is not warranted.

Would the AIC and gun controllers have us believe that the large peak for '87 and '88 for Victoria was caused by a huge peak in firearm possession rates in Victoria during that time?  Firearm possession rates don't rise and fall so abruptly, so something else was responsible for the '87-'88 peak.

The large '87-'88 peak for Victoria was caused by a combination of a large increase in suicide (to 1.1) for '87 followed by a large peak in homicide (to .8) for '88.  About the only thing that suicide and homicide have in common is desperate people.  Maybe a year of people giving up in desperation was followed by a year of people striking out in desperation.  Or, maybe the two back-to-back peaks were entirely random.

A gun controller would probably like to think that there is some positive significance in the last parts of the plots for NSW and Victoria.  Both are almost identical except that the last peak followed by steady decline began a year earlier for Victoria (peak in '91) than for NSW (peak in '92).  The start of the decline for NSW was in '93, three years after effectivity of the new NSW Firearms Act of 1989 but just one year after effectivity ('92) of changes made in that law (in '91) as a result of the "select committee on gun law reform."  It makes no sense that such an abrupt change would result from something that began three years earlier, so the basic NSW Firearms Act 1989 had no apparent impact.  But, could the '91 amendments to the act have caused the downturn in '93 & '94?

There are two things to consider in the search for the answer to this question.  First, two years don't make a trend.  As we've already seen, it is entirely possible to get two years of "trend" just from random variations of the components of "gun deaths" coming together occasionally.  The second thing to consider is the makeup of the numbers.

Examination of the death rate data at NISU reveals that the '93 & '94 declines for NSW resulted from two small declines in the suicide rate from the extraordinary peak in '92, combined with an extra high accident rate for '92 followed by a drop (to virtually zero) for '93 and '94.  Gun homicide was up a little in '93.  So, the two years of decline were caused by a combination of ups and downs that could have easily been random.

NSW Gun Death Rates, 
Figure 4. NSW Gun Death Rates,
Post Firearms Act
The matter of there being only two years of trend can be resolved now by actually looking at whether or not the "trend" continued up until the laws got changed again in '96.  Now we can see that the downward drops in '93 and '94 wasn't a trend at all.  After the steep drop through '93 and '94, the NSW gun death rate nearly levelled off for '95-'97 as first homicide rose a little in one year followed by the accident rate rising in the next.  However, after the initial steep drop for two years, the rate did generally continue to drop although much more slowly (.18 in 3 years) basically because the gun suicide rate continued to drop slowly through '96.

Although the steep dropping wasn't a true trend, in that it lasted only two years, could it have been that the descent was just a transition to a new lower rate caused by the '91 amendments to the law?  Although it does not bear upon the question of whether or not the AIC was justified in claiming that two years of death rate decline could be attributed to new gun laws, the steady decline in suicide did in fact actually extended through '96.  Was there anything in the '91 amendment that could have caused sustained reductions in suicide?

Gun Death Rates, Victoria, '91-'97
Figure 5. Victoria Gun Death Rates,
Post Firearms Act Amendment
The situation is different for the downward trend at the end for Victoria.  The trend was almost entirely the result of a downward trend in gun suicides, plus small, random, increases for gun homicide for '92 and '94.  We find now that the trend gradually levelled off only because of random bobbling up and down of the gun homicide and gun accident rates, since the gun suicide rate abruptly leveled off from '94 for the remainder of the time until the Victoria gun laws were changed again in reaction to the '96 Port Arthur massacre.  If a rise or drop is part of a trend to a new level, the transition from the rise or drop to the new level is generally gradual, not abrupt. Therefore, the levelling off of the suicide rate was probably caused by something different than the initially dropping rate.  That is, it is most likely that the dropping and the levelling off were two separate trends caused by two different things.

Note that the plots of different kinds of rates for different states have some features that could be informative if we had some information about other things that happened at various times, so that we could see that such things might correlate with gun death rates.  For example, did Victoria do something to make some of its citizens desperate in '87 and '88?  Note too that combining the data for all of Australia, rather than examining the data for each state separately, would dilute the data corresponding to the features visible in the plots, and would thereby obscure relationships (i.e., cover up the readily apparent state plot features) since all the different jurisdictions had different laws, etc. at different times.  When doing time series statistical analysis, it is most productive to use data points for the largest groupings that have the same things (like rules or processes going into effect) happening to the entire grouping.  This restriction of having the same things happening throughout the grouping results in reduced grouping size, like to single states, which works fine as long as the populations of the smaller groupings are sufficient to yield the necessary statistical significance.